Profile My skills Saved Collections Edit profile Submit a skill

Peft Rlvr Evaluation

Comprehensively evaluate 12+ parameter-efficient fine-tuning methods for RL with Verifiable Rewards (RLVR). Show DoRA/AdaLoRA outperform LoRA, SVD-based methods fail on RL, extreme reduction creates bottlenecks—providing empirical evidence that geometric-aware adapters align better with RL's off-principal update dynamics.

adu2021 38c6c10 993 B Updated
Coding & Dev Tools

File contents

Overview

Reveals LoRA's limitations in RL contexts through systematic PEFT evaluation.

Core Technique

Structural Variant Comparison:

# Standard LoRA sometimes suboptimal
methods = [LoRA, DoRA, AdaLoRA, PiSSA, VeRA, ...]

# Evaluate on RLVR (sparse binary rewards)
results = benchmark_all_methods(methods, rlvr_training)

# Finding: DoRA/AdaLoRA outperform

When to Use

Use when: RL fine-tuning, parameter efficiency critical, method selection.

References

  • Comprehensive PEFT method evaluation
  • RL-specific optimization dynamics
  • Geometric-aware adapter superiority

adu2021/skillxiv/tree/main/skills/skillxiv-v0.0.2-claude-opus-4.6/peft-rlvr-evaluation commit 38c6c10978

Browse category

Frequently asked questions

npx skillmds@latest add adu2021/peft-rlvr-evaluation